A better design would be to explicitly acknowledge that words have
meanings. The way that this is currently done looks roughly like this:
(EvaluationLink
(PredicateNode "is")
(ListLink
(ConceptNode "apple@meaning-42")
(ConceptNode "fruit@meanning-66")
)
)
I hope the above is "obvious": the 42nd kind of meaning of the word "apple"
is a kind of "fruit", where by "fruit", we mean the 66th entry in Webster's
dictionary.
(ReferenceLink
(ConceptNode "apple@meaning-42")
(WordNode "apple")
)
That tells you the actual word that gets used for meaning-42. This is a
lexical function https://en.wikipedia.org/wiki/Lexical_function
(WordInstanceLink
(SentencNode "id1")
(WordInstanceNode "apple@bf71826c-487e-42df-a941-0ecd3c942a76")
)
This tells you that the the word apple occurred in sentence id1
(ReferenceLink
(WordInstanceNode "apple@bf71826c-487e-42df-a941-0ecd3c942a76")
(ConceptNode "apple@meaning-42")
)
This tells you that the word apple in sentence id1 actually corresponds to
meaning 42.
See? No context link at all.
The above oversimplifies things a little bit. Some of the reference links
should probably be EvaluationLinks. The lexical functions need to be
improved, a lot. The current output is documented here:
http://wiki.opencog.org/w/RelEx_OpenCog_format but it could be over-hauled
and improved, its not perfect.
I believe that the above should work well with PLN, but that remains to be
seen: again Nil is working on this now.
--linas
On Mon, Nov 14, 2016 at 9:34 AM, Vishnu Priya <[email protected]>
wrote:
> Hey Linas!
>
> Thanks for the reply. It's ok. Totally understandable.!!
>
> Yeah just read about ContextLink on wiki.
>
> I have a scenario, where i have sentences that i want to give to NLP
> Pipeline. Along with sentences, i also have an attribute called id. Like a
> reference for sentence.
> Each sentence is associated with an identifier. For me, it would be useful
> when i have the sentences parsed along with their id.
> Later, say i stimulate and get STI, whatever i do, finally i should be
> knowing, to what id the atom belongs to.
>
> So i thought, with something like below, i might achieve that.
> apple is fruit in the context of id1.
> (EvaluationLink
> (ContextLink id1
> (PredicateNode "is")
> (ListLink
> (ConceptNode "apple")
> (ConceptNode "fruit")
> )
> )
>
>
>
> But i don't know, how to input my sentences along with their identifiers.
> Is it possible somehow to do such a thing of incorporating identifiers ?
> or is it totally not doable?
>
>
> --vishnu
>
>
>
>
> On Friday, 11 November 2016 02:30:44 UTC+1, linas wrote:
>>
>> Hi,
>>
>> sorry just now recovering from system outages and an email overload.
>>
>> ContextLink and how to use it is documented on the wiki.
>>
>> Currently it it not used very much, or at all.
>>
>> ContextLinks only make sense once you know how to asssign meaning to
>> things -- syntax parsing of sentences is far too low-level for this,
>> because you don't yet know what the word "apple" is.
>>
>> --linas
>>
>>
>>
>> On Fri, Oct 21, 2016 at 10:27 AM, Vishnu Priya <[email protected]>
>> wrote:
>>
>>> Hey Linas,
>>>
>>> I would like to know how to use ContextLink.
>>>>
>>>
>>> - The Apple is red in color.
>>> - The Headquarters of apple is in California.
>>>
>>> Each and every sentence of mine has certain context word.
>>> I want the former sentence to be parsed along with ContextLink fruit
>>> and the later as company. So that later, i can identify which atom belongs
>>> to which context.
>>> Should i make changes at the parser level? What should i do?
>>>
>>> Cheers,
>>> Vishnu
>>>
>>
>>
--
You received this message because you are subscribed to the Google Groups
"opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/opencog.
To view this discussion on the web visit
https://groups.google.com/d/msgid/opencog/CAHrUA35T%2BiU0M0d_zT%2BirOLVOmnMpaaqVrC7UZK14YS0CWz53A%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.
